Predicting the fitness of mutations in the evolution of pathogens is a long-standing and important, yet largely unsolved problem. In this study, we used SARS-CoV-2 as a model system to explore whether the intrahost diversity of viral infections could provide clues on the relative fitness of single amino acid variants (SAVs). To do so, we analysed ~15 million complete genomes and nearly ~8000 sequencing libraries generated from SARS-CoV-2 infections, which were collected at various timepoints during the COVID-19 pandemic. Across timepoints, we found that many successful SAVs were detected in the intrahost diversity of samples collected prior, with a median of 6-40 months between the initial collection dates of samples and the highest frequency seen for these SAVs. Additionally, we found that the co-occurrence of intrahost SAVs significantly captures genetic linkage patterns observed at the interhost level (Pearson's r=0.28-0.45, all p<0.0001). Further, we show that machine learning models can learn highly generalisable intrahost, physiochemical and phenotypic patterns to forecast the future fitness of intrahost SAVs (r2=0.48-0.63). Most of these models performed significantly better when considering genetic linkage (r2=0.53-0.68). Overall, our results document the evolutionary forces shaping the fitness of mutations, which may offer potential to forecast the emergence of future variants and ultimately inform the design of vaccine targets.
